An Enhanced Framework for Detecting Fraudulent Apps using Sentiment Analysis and User Review Analysis

Abstract: Malicious mobile applications pose a significant threat to users by simulating the appearance and functionality of legitimate software. Upon installation, such applications often generate revenue through intrusive advertising, unauthorized data extraction, malware distribution, and other harmful activities. The primary challenge lies in end-users' inability to reliably distinguish between authentic and counterfeit applications—a gap that has led many users to increasingly rely on peer-generated reviews as a decision-making tool prior to installation. This research proposes a comprehensive information platform designed to facilitate informed user decision-making through aggregated user reviews and ratings that reflect genuine user experiences with specific mobile applications. To enhance the utility of such reviews, this study employs sentiment analysis techniques to classify textual feedback according to its emotional valence, thereby categorizing user opinions as positive, negative, or neutral. This methodological approach aims to provide users with a nuanced understanding of application quality and reliability.

Keywords: User Reviews, Sentiment Analysis, Lexicon-Based Analysis, Tokenization, Stop-Word Removal.